The Design and Engineering of Curiosity

by Emily Lakdawalla

  The rocker-bogie suspension system actually performs better rolling bogie-first than rocker-first. Curiosity, the Mars Exploration Rovers, and Sojourner have all been designed to drive rocker-first so that if a forward drive gets the rover into a hazardous situation, it is more likely to be able to back straight out of the problem terrain. Curiosity sometimes drives backwards, but when facing backward the MMRTG obstructs the Navcams’ view of the nearby terrain, preventing the rover drivers from obtaining the images they need to plan future drives. So backwards drives have to finish with a turn or at least a wiggle to one side or the other to allow the Navcams to see the terrain ahead.

  4.6.2 Motors

  Curiosity’s actuators consist of a motor, a gearbox, a brake, and an encoder; in this book, “motor” typically applies to a whole actuator assembly (Figure 4.12). The motors are very powerful. A single drive motor has enough torque to drive the rover up a vertical wall. The rover’s top speed, 4.2 centimeters per second (151 meters per hour), is so slow that the motion is quasi-static. There is no freewheeling, and all wheel rotation is commanded wheel rotation. When the wheels aren’t rolling, they are braked.

  Figure 4.12. One of the high-torque drive motors for Curiosity’s mobility system. The motor end is at the right side; its output passes into a four-stage gearbox that rotates the plate at left. From Cook (2009).

  Because only the four corner wheels are steerable, the rover can’t “crab” (drive sideways), but it can turn in place, allowing it to pick its way safely among a field full of obstacles provided that the obstacles are separated by more than the width of the rover. The steering motors are positioned above the wheel’s centers, connected by U-shaped brackets to the motorized wheel hubs, so that the wheels steer in place about a vertical axis. A turn in place of 60° or more draws a complete circle of wheel tracks on the ground, leaving telltale “donuts” about 2.75 meters in diameter along the rover’s tracks (Figure 4.13).

  Figure 4.13. Donuts along tracks document rover turns. Top: Right Navcam photo taken after a drive on sol 527 (NRB_444289916RADLF0260000NCAM00252M1) showing marks of two turns in place: “B” is a complete donut, and “A” is not. Credit: NASA/JPL-Caltech. Bottom: Mars Reconnaissance Orbiter HiRISE image ESP_035350_1755 taken sol 538, including donuts A and B as well as tracks of several arcing turns. NASA/JPL-Caltech/UA/Emily Lakdawalla.

  The rover’s motor controller can only run eight motors at a time, so the rover cannot steer and drive simultaneously. Thus drives alternate between straight drive segments and arcing turn segments. Wheel rotation rates are adjusted for arcing turns so that the wheels on outer edges of turns rotate faster than inner wheels.

  4.6.3 Wheels

  Curiosity’s wheels presented a design challenge because they had to serve as both landing gear and running gear.37 As landing gear, they had to absorb the mechanical shock of touchdown, protecting the wheel motors from harm. After landing, the wheels needed to provide good traction over Martian terrain, including floating the heavy rover over sand. They needed to be as lightweight as possible, and to fit within the narrow confines of the aeroshell. The final design represents a compromise among all these competing requirements. Surviving the landing trumped all other requirements, and most of the design effort focused on ensuring Curiosity could drive away from any imaginable landing scenario. Unfortunately, that proved shortsighted.

  The wheels are 50 centimeters in diameter at their centers (including the height of the treads), with a crowned profile such that they are 46.5 centimeters in diameter at their outer edges (Figure 4.14). They are 40 centimeters wide. They consist of an aluminum tire and a titanium hub-and-spoke assembly. The spokes have a complex shape that makes them springy in all directions, allowing them to do their job of absorbing a landing jolt even if they landed on slopes or rocks.

  Each wheel was machined from a single block of aluminum. The wheel skin is incredibly thin – at just 0.75 millimeters, as thin as it was possible to machine – in order to limit the wheels’ total mass. The wheels are stiffened by three circumferential rings: two at the inner and outer edges, and a third ring located about a third of the way inside the outer ring to provide a place for the spokes to attach. Together, all these design elements enabled the wheels to deform dramatically under the force of a landing and return to their original shape (Figure 4.15).

  Figure 4.14. Parts of Curiosity’s wheel. Curiosity wheels are crowned, 50 centimeters tall at center, 46.5 centimeters diameter at sides, and 40 centimeters wide. NASA/JPL-Caltech/Emily Lakdawalla.

  Figure 4.15. Spring-like deformation of a rover wheel during testing. In this test, two of the spokes have “bottomed out” on the inside surface of the wheel. After this test, the wheels sprang back to their original shape. From Lee ( 2012 ).

  Other design elements had to do with surface operations. A black anodized coating prevents the wheels from throwing glints into camera images. For traction, the wheels have treads or “grousers”. The height and spacing of the grousers represent a compromise among several factors. The grousers had to be spaced close enough that they would cog with features on rock faces, about 65 millimeters apart. Their height is relatively short. Through laboratory tests of different tread designs, the mobility team found that most of the improvement in wheel traction came with treads whose height was comparable to the particle size of the material the wheel drove on. They settled upon a tread height of about 3% the wheel radius, or about 7 millimeters. After the challenging Opportunity experience of driving a rover across sloping crater walls, in which the rover tended to skid downslope, they added a chevron pattern to the wheel treads in order to prevent the same from happening to Curiosity.

  If the ground were perfectly flat and rigid, the crowned shape of the rover wheels would touch it only at one point. In reality the weight of the rover drives it in to the ground, so to approximate the ground pressure of rover wheels on the surface, engineers defined the contact area as being the wheel width times the wheel radius. (This effectively assumes that 57° of the wheel’s full circumference is in contact with the surface.) In operation, the wheels do not generally touch the ground over so much of their radius (Figure 4.16).

  Figure 4.16. Wheel performance on different substrates. Upper left: small rocks over regolith, the substrate encountered by most previous missions. The wheels dig slightly into the surface, but only a small area of the wheel is in contact. Upper right: a jagged, rocky surface. At times, the rocks contact the surface at only one point, as the right rear wheel does here. The wheel skin is thick enough that the rover’s weight merely resting on a pointy rock does not puncture a wheel. Lower left: a well-packed sand ripple on which the wheels are getting good traction, similar to that in the upper left image. Lower right: a ripple made of fluffier sand into which the wheels are embedding as they slip. MAHLI images 0504MH0002610000200627E01, 0506MH0002610000200672E01, 0529MH0002610000201142E01, and 0711MH0002610010204346E01, NASA/JPL-Caltech/MSSS.

  The wheels have twelve holes cut into them, part of an asymmetric tread feature that interrupts the otherwise regular pattern of the wheel treads (Figure 4.17). This feature makes marks at regular intervals (about 1.5 meters apart) in rover wheel tracks. The track markings can be directly compared to the expected distance traveled in order to measure how much the rover wheels have slipped during a traverse. In initial wheel designs, these features were the letters “J P L” machined into the pattern of the treads (see Figure 4.11), but NASA objected to JPL labeling the wheels in this way. So the design was changed to one that incorporated bland rectangular holes. Mischievously, the wheels’ designers made those holes spell out “J P L” in Morse code in the tracks.

  Figure 4.17. Asymmetric tread features in the rover wheels mark the rover’s tracks every time the wheel has rotated once, about once every 1.5 meters. They also spell out “J P L” in Morse code. J is · - - - ; P is · - - · ; L is · - · · . Mosaic of two left Navcam images taken on sol 535. NASA/JPL-Caltech.

  4.6.4 Wheel degradation />
  The wheels performed perfectly upon landing; the only visible damage from the landing event was a tiny crack in the left middle wheel. During the first 500 meters of the rover traverse from Bradbury Landing to Yellowknife Bay, the wheels suffered little additional damage. The team surveyed the wheels with MAHLI on sol 411, noticing a puncture in the left front wheel. Re-imaging the wheels on sol 463, they observed that the tear had grown dramatically worse. From that sol forward, the team commanded numerous wheel-imaging sequences, shooting photos of all wheels with the MAHLI camera and the right-side wheels with Mastcam in between every drive. Periodically, they would devote an entire drive to full surveys of the wheels by driving the rover short distances between four or five wheel surveys in order to image the entire wheel surface. Wheel imaging is summarized in Box 4.1.

  Box 4.1. Sols with MAHLI and ChemCam RMI wheel imaging to sol 1800.

  Over time, more punctures and tears appeared in the middle and front wheels, while the rear wheels remained relatively unscathed. Figure 4.18 through Figure 4.22 document the condition of all six wheels at three points in the mission. As of sol 513, the rover had driven 4.7 kilometers; as of sol 708, 8.7 kilometers; and as of sol 1513, 15.1 kilometers. Improved understanding of how to save the rovers’ wheels slowed the progression of damage after sol 708. Fewer new punctures formed, but dents and cracks progressed. The first wheel images to reveal broken grousers, on the left-middle wheel, were taken on sol 1641, after about 16 kilometers of driving. As of sol 1800 the rover has driven 17.5 kilometers, with no further broken grousers.

  Figure 4.18. MAHLI survey of right front wheel on sol 513 (left column), 708 (middle), and 1313–1315 (right). See text for discussion. NASA/JPL-Caltech/MSSS.

  Figure 4.22. MAHLI survey of left front wheel on sol 513 (left column), 708 (middle), and 1313–1315 (right). See text for discussion. NASA/JPL-Caltech/MSSS.

  Figure 4.19. MAHLI survey of right middle wheel on sol 513 (left column), 708 (middle), and 1313–1315 (right). See text for discussion. NASA/JPL-Caltech/MSSS.

  Figure 4.20. MAHLI survey of rear wheels on sols 513 and 1313–1315. See text for discussion. NASA/JPL-Caltech/MSSS.

  Figure 4.21. MAHLI survey of left middle wheel on sol 513 (left column), 708 (middle), and 1313–1315 (right). See text for discussion. NASA/JPL-Caltech/MSSS.

  The punctures were caused by two factors: metal fatigue and forces intrinsic to the rocker-bogie suspension system.38 Fatigue is a consequence of the flexibility of the wheels. They flex back and forth with each wheel rotation. Stress concentrates at the tips of the chevron shapes in the grousers. Eventually the skin cracks near chevron points, and over time the cracks grow and merge. Once the cracks propagate entirely across the width of the wheel, the grousers are unsupported by skin stretching between them, so they flex even more with each wheel rotation. Eventually, repeated flexing fatigues the grousers and they also begin to snap.

  At first, it was difficult to understand why only the left and middle wheels seemed to be getting damaged. When at rest on level ground, the front, middle, and rear wheels on each side bear weights of 564, 636, and 458 newtons, respectively. Experiments in the Mars Yard showed that wheels were robust to punctures at these forces. It takes 800 newtons for a sharp metal cone to puncture wheel skin, and 1500 newtons for sharp rocks collected from the Mojave Desert to do so.

  How could such high forces on wheel surfaces be generated? Earth tests revealed that the answer lay in kinematics of the full rocker-bogie suspension system. The motors drive all six wheels at fixed rotation rates. When one wheel encounters an obstacle that does not move aside or press into the ground under the weight of the rover, that wheel must travel a longer distance than the other wheels as it rolls over the obstacle. But its motor drives it at the same fixed rate as the other wheels. So the wheel encountering the obstacle is dragged as it travels.

  Additionally, when the rover drives forward, four of its six wheels are on forward-projecting legs. If one of these wheels climbs an obstacle, some of its rotation rate goes toward vertical motion, and its horizontal motion is slowed. The remaining wheels continue moving horizontally at full speed, shoving the blocked wheel on its forward-projecting leg toward the obstacle with considerable force. If the obstacle is strongly cemented into the ground and pointy, it can open a hole. Consider a rolling suitcase: when you drag it behind you, you exert an upward force as you pull on the wheels, helping it to climb an obstacle. When you push it in front of you, you exert a downward force as you push, and the suitcase’s motion is easily and often stopped by small obstacles, jarring your arm. In tests in the Mars Yard of driving forward over sharp cones, the front and middle wheels punctured easily, while the rear wheels remained whole.

  The mission dramatically reduced the rate of damage by:Picking local drive paths carefully among potentially damaging rock, consequently ending the use of autonav (which would blithely drive the rover over pointy rock patches).

  Mapping the terrain ahead using orbital data (including HiRISE images and Odyssey THEMIS thermal inertia maps) and seeking out less “pointy” terrain during long-term traverse planning.

  Avoiding turns on sharp terrain.

  Sometimes driving backwards.

  After sol 660, the engineers decided that the turns-in-place required at the ends of backwards drives in order to do drive-direction imaging held more potential for wheel damage than was saved through backwards driving.

  As of the mission’s second landing anniversary, when the rover had driven about 8 kilometers, the engineers estimated the following remaining lifetime for the wheels:39 Bedrock with lots of rocks: 8 kilometers.

  Lots of rocks, not on bedrock: 13–14 kilometers.

  Bedrock with few rocks (like flagstones): 30–40 kilometers or more.

  Smooth or sandy, with few or no rocks: indeterminate (causes no damage).

  Furthermore, Mars Yard testing suggests that, on average, once three grousers have broken on a wheel, about 60% of its life has been consumed.40 The rover’s wheels are now expected to survive as long as the mission does, although they may look much the worse for wear by the time the mission ends. Curiosity should be able to achieve at least 28 kilometers total mission odometry unless there is a dramatic change in the terrain.

  On sol 1646, in response to the observation of broken grousers on sol 1641, the mission tested new traction control ability for the first time.41 Traction control was turned on by default on sol 1678. The rover senses when a wheel is climbing an obstacle by monitoring tilts of rockers and bogies. The rover responds by slowing the turn rate of the wheels that are not climbing obstacles, allowing the climbing wheel to rotate faster, thereby reducing the likelihood of punctures and widening cracks.

  Even with “failed” wheels the rover may continue to be able to drive. The wheels fail when all the grousers have snapped, leaving the inner two-thirds of the wheel diameter flapping, connected to the rest of the wheel only at the locations of the asymmetric tread features (Figure 4.23). This is hazardous to the rover, because sharp edges on the broken wheels can scrape against the cable that runs to the wheel motors. Slicing into a cable could not only jeopardize the functioning of that wheel’s motors, it could also potentially cause a short circuit that would risk the motor controller – which also controls the motion of all other moving parts on the rover. Driving the rover with a wheel in this condition on Mars could be hazardous, but it would still be better than not driving at all. In the Mars Yard, driving on such wheels has been tested; eventually the inner two-thirds of the wheel snaps off completely, and the rover is able to drive quite effectively on the remaining third of the wheel surface that is still attached to the inner stiffening ring.42

  Figure 4.23. Wheel tested to failure. The rover can still drive effectively on this wheel, but the sharp edges of the broken grousers and webbing present a hazard. Photo taken in the JPL Mars Yard on October 13, 2014 by Emily Lakdawalla.

  4.7 TESTBEDS

  4.7.1 The Mars Yard

&n
bsp; A rover as large as Curiosity requires a large area for testing purposes. The JPL Mars Yard is 66-by-36 meters in size, located at the top of the steep Pasadena campus (Figure 4.24).43 Most of it is flat and level, with the surface material made of beach sand, decomposed granite, brick dust, and volcanic cinders. There are also lots of basalt rocks of different sizes that engineers can move around to simulate different driving conditions. One side of the Mars Yard is sloped at a range of angles for testing driving and arm operations on sloping surfaces. At one end is a small building that garages test rovers, associated equipment, and engineers (Figure 4.25).

  Figure 4.24. Panoramic view of the Mars Yard at JPL. NASA/JPL-Caltech.

  Figure 4.25. MAHLI self-portrait of the vehicle system testbed taken inside the Mars Yard shed, August 1, 2012. NASA/JPL-Caltech/MSSS.

  4.7.2 The Vehicle System Testbed

  The Vehicle System Testbed (VSTB), also known as “Maggie,” is the highest-fidelity copy of the rover and is housed in the shed at JPL’s Mars Yard.44 It is used for testing driving, arm movements, and drilling using the same software and electronics that are on Mars, on a suspension system that will put the rover in similar positions as experienced on Mars.45 It has the same body, suspension system, arm, sample handling system, mast, and other motorized elements as the flight rover. Initially, it had the same wheels as the flight rover, but after degradation they were eventually replaced with wheels twice as thick as those on Mars. (Their rapid degradation resulted in part from bearing nearly the full Earth weight of the full-scale rover.)